A polarization fusion network with geometric feature embedding for SAR ship classification

作者：

Highlights：

• A PFGFE-Net is proposed for SAR ship classification. It offers an effective solution to combine dual-polarization features and geometric ones.

• Besides the fundamental dual-polarization features, the polarization coherence feature is applied to expand network inputs to improve the accuracy.

• PF is achieved from the input data, network feature-level, and output decision-level.

• GFE is proposed to enrich model expert experience. It can give impetus to the CNN classification model to further improve the accuracy.

摘要

•A PFGFE-Net is proposed for SAR ship classification. It offers an effective solution to combine dual-polarization features and geometric ones.•Besides the fundamental dual-polarization features, the polarization coherence feature is applied to expand network inputs to improve the accuracy.•PF is achieved from the input data, network feature-level, and output decision-level.•GFE is proposed to enrich model expert experience. It can give impetus to the CNN classification model to further improve the accuracy.

论文关键词：Synthetic aperture radar (SAR),Ship classification,Convolutional neural network,Polarization fusion (PF),Geometric feature embedding (GFE)

论文评审过程：Received 26 July 2021, Revised 27 September 2021, Accepted 3 October 2021, Available online 9 October 2021, Version of Record 16 October 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108365